基于灰度共生矩阵信息融合的视频噪声识别算法

摘要

For the noise in video monitoring images,we propose an objective recognition and evaluation method for noisy video with five features information fusion and on the basis of gray level co-occurrence matrix (GLCM)by making use of difference information between two adjoining frames in video which can reflect the noises in video and combining GLCM.The method first conducts differential operation on two consecutive frames of video,and then applies GLCM to frame difference image to calculate five features of the image including information entropy,energy,inverse gap,sum of mean,and sum of variance.Finally,it uses support vector machine for training and learning of the information fusion to effectively identify the noisy video without reference.Experimental results show that and the prediction and recognition effect of this method is quite good,the recognition rate achieves up to 95.11% with fairly good consistency with the subjective recognition, and can get lower missing rate and false positive rate.%针对视频监控图像中存在噪声的问题，利用视频中相邻两帧间差异信息可反映出视频中的噪声，结合灰度共生矩阵，提出基于灰度共生矩阵五特征信息融合的客观噪声视频识别评价方法。该方法首先对连续的两帧视频进行差分运算，然后对帧差图像利用灰度共生矩阵计算图像信息熵、能量、逆差距、均值和、方差和五个特征，最后用支持向量机进行信息融合训练学习，对无参考视频噪声有效识别。实验结果表明，该方法预测识别效果较好，识别率高达95．11％，与主观识别有较好的一致性，能获得较低的漏检率和误报率。